Computational Protein Design: Regenerative Medicine's Healing Future

Today, I’m excited to cover two of the most science fiction-sounding fields out there: computational protein design meets regenerative medicine! Let’s jump into it.

Protein Design Meets Bone Repair

Regenerative Medicine: Real-Life Sci-Fi

Every time I think about regenerative medicine, I can only picture the scene in “The Empire Strikes Back” when Luke gets a robotic hand that attaches to his arm.

And this is not a bad thing! This sci-fi sounding field focuses on healing or replacing tissues and organs by using the body’s natural processes, or engineering new ones. It includes a wide range of approaches:

• Stem cell therapies

• Gene therapies

• Tissue engineering

• Biomaterials and scaffolds

In the clinic today, biomedical implants are the most common regenerative technologies. Think orthopedic screws, dental implants or hip replacements. Unfortunately, most of the materials used don’t communicate well with the body. Most are inert, failing to promote sufficient cell attachment for tissue integration or healing.

To make better implants, we need better coatings: bioactive materials that talk to cells and guide healing. Real functional materials!

The Cell’s Perspective: Fibronectin and Integrins

Okay, let’s take a step back and look at the biology of things.

In healthy tissues, cells live on and respond to the extracellular matrix. One of its main components is fibronectin. This glycoprotein plays a key role in cell adhesion, growth, migration, and wound healing: a little bit of everything!

To do its job, fibronectin binds to cell-surface receptors, especially the integrin 𝜶5𝜷1, which helps cells attach and move. This makes 𝜶5𝜷1 vital for the development and regeneration of tissues.

Fibronectin binds to 𝜶5𝜷1 using a small loop called RGD, which can also be produced on its own as a peptide. People have tried to use fibronectin and the RGD peptide in medical materials, but they have problems:

• Fibronectin is extracted from human plasma, so the production is not scalable, and the product is very delicate

• RGD, while easier to produce and purify, has a much weaker effect on cells

Ideally, the solution is a protein that binds like fibronectin, but is also easy to make and stable!

Computational Protein Design to the Rescue

The authors of today’s paper used computational protein design to solve this real-world problem.

The researchers realized that 𝜶5𝜷1 has a unique pocket they could target. The team got busy designing new protein binders, choosing the small protein ferredoxin as a stable scaffold. They explored 2 strategies:

1. Starting with the RGD loop, they rebuilt the protein scaffold around it, fragment by fragment

2. Building the scaffold first, then adding the RGD loop.

The team used Rosetta to optimize the sequence and the structure in silico. Their aim? To obtain the most shape complementarity to 𝜶5𝜷1. The researcher selected over 20,000 designs to test experimentally! 8000 from the first approach and 12000 from the second. Crazy numbers!

From the Computer to the Lab: Testing the Designs

Using yeast surface display, the team screened thousands of designs and narrowed them down to 16 promising candidates.

But they were still not satisfied.

From these 16 designs, they generated site saturation mutagenesis libraries, systematically mutating each amino acid to improve binding. This huge library of mutants was screened against 𝜶5𝜷1 and a closely related integrin to fine-tune specificity. Out of these, 5 had great affinities for 𝜶5𝜷1, in the nanomolar range!

NeoNectins: New, Tight Binders for 𝜶5𝜷1

The researchers baptized the new binders NeoNectins, inspired by fibronectin.

Their most promising binder was NeoNectin candidate 1, or just NN-C1. Besides the not-so-great name, the performance is amazing:

• NN-C1 binds 𝜶5𝜷1 600x better than fibronectin

• And a silly 167,000 times better than RGD!

• There is no cross-reactivity with other integrins, in vitro or in cells

The team tested the molecular basis of the interactions using electron microscopy. When activated, 𝜶5𝜷1 shows dramatic conformational changes, turning into an extended, open conformation. NN-C1 stabilizes this open conformation, just like (or even better than) fibronectin, explaining the amazing performances on cells.

And the protein structures also showed the expected interactions between NN-C1 and 𝜶5𝜷1: a beautiful validation of the computational design!

Functional Testing: Soluble and Immobilized NN-CN1

Getting confirmation of the binding is cool, but the researchers were interested in the actual functionality of NN-C1.

First, they tested the soluble form. When used in solution, NN-C1 blocked fibronectin-dependent adhesion and migration, just as expected. Cells treated with soluble NN-C1 spread less, moved more slowly, and expressed fewer adhesin-related genes.

Then, they studied the immobilized form. The team grafted NN-C1 onto hydrogels, soft, water-based materials used as 3D scaffolds for stem cells. Compared to fibronectin and RGD, the cells with NN-C1 supported better cell spreading, survival, and general health. NN-C1 could be an ideal component for tissue engineering!

Next, they coated titanium discs with NN-C1, fibronectin, or RGD. Titanium is one of the most used materials for implants, thanks to its biocompatibility, strength, and ability to integrate with bone. NN-C1 matched or outperformed fibronectin and just destroyed RGD. And cells grown on NN-C1 and fibronectin showed similar gene expression, suggesting that cells “see” them in the same way.

Testing NN-C1 In Real Implants

Okay, so this is the coolest part of the paper: real-world potential!

The team evaluated the efficacy of NN-C1-grafted implants in vivo, in a rabbit model (I am not a fan of animal testing, but unfortunately, it is still necessary in cases like this). After inserting implants grafted with NN, RGD, FN, or just bare titanium, they evaluated the responses. And the results were great:

• Bone Growth: NN-C1 implants led to the highest bone volume/total volume ratios at both 3 and 6 weeks post-surgery.

• Bone Quality: Histology and SEM revealed denser, more structured bone around NN-C1 implants.

• Bone-Implant Contact: BIC values were highest for NN-C1, which was the only condition that maintained or improved integration over time.

• Biocompatibility: The team didn’t see any adverse effects, confirming the safety profile of NN-C1!

Conclusions and Future Direction

NeoNectins show a promising path for any biomaterials involved in tissue regeneration:

• It’s small, just 65 amino acids, stable, and easy to manufacture

• It binds 𝜶5𝜷1 with high specificity and affinity

• It promotes tissue regeneration as well or better than fibronectin, working both in vitro and in vivo!

But this work also shows the potential of computationally generated proteins for regenerative medicine (and not only!). This approach to grafting small, stable proteins to biomaterials could also be extended to encapsulating therapeutics or to recruiting specific cells for tissue healing!

Don’t miss anything and read the whole paper here!

If you made it this far, thank you! What do you think about this? Do you see a future for regenerative medicine? What do you think are the biggest problems? Reply and let me know!

More Room:

• AI’s Meeting with DNA Origami: Can you put together DNA origami, TEM and AI? Of course you can! And it’s even useful. This study shows that AI, using CNN models, can accurately analyze DNA origami nanostructures in TEM images. After training on simulated and experimental data, the VGG16 model proved most effective for quickly determining ligation numbers, offering a reliable tool for characterizing DNA nanostructures. It would have helped me a few years ago!

• Directing DNA switches: DNA computing is a pretty cool area. This study presents a new method for building DNA logic gates on DNA origami using directional strand polymerization. DNA switches on the origami surface trigger a controlled chain reaction in response to input strands, enabling precise, localized logic operations. Fluorescent reporters allow in situ, single-molecule observation of these processes. Good idea for high-density computation!

• Meta Superlattices: No, I am not talking about Mark Zuckerberg. Meta-DNA extends DNA’s programmability to bigger structures. This study presents a method for assembling 1D porous superlattices using DNA-modified meta-DNA. By controlling DNA bond placement, the team created tunable meso- to macroporous structures with enhanced electrocatalytic performance, demonstrating the potential of DNA-based porous materials for energy and catalytic applications.

• Know someone who’d love this?
  Pass it on! Sharing is the easiest way to support the newsletter and spark new ideas in your circle.

• Got a tip, paper, or topic you want me to cover?
  I’d love to hear from you! Just reply to this email or reach out on social.